A Study on Evolutionary Reactive Ordering Systems

为了开发面向多阶段生产系统的敏捷订货系统，本研究提出了一种缓冲库存反应式控制器，以应对需求的不稳定变化。在控制器中，不仅可以检测到需求的不稳定变化，而且可以检测到需求的均值和方差，并使用指数加权移动平均(EWMA)图确定缓冲库存的最优水平。除了使用EWMA图的控制器外，还针对两种类型的准时制(JIT)订货系统，看板系统和并行订货系统，提出了一种基于神经网络的控制器。通过各种不稳定需求条件下的仿真实验，验证了该系统的有效性。仿真实验结果表明，所提出的控制器不仅能对需求的均值变化做出反应，还能对需求的方差变化做出反应，满足产品需求平均等待时间的要求。此外，通过使用EWMA图可以看出，基于神经网络的控制器比基于EWMA图的控制器显著降低了系统的平均在制品库存。

In order to develop an agile ordering system for multi-stage production system, this research proposed a reactive controller of buffer inventory for reacting to the unstable changes in demand. In the controller, the unstable changes in demand, not only the mean but also the variance, can be detected, and the optimal level of buffer inventory is determined as a reaction to the changes using the exponentially weighted moving average (EWMA) charts. In addition to the controller using EWMA charts, a controller using a neural network is proposed for two types of just-in-time (JIT) ordering systems, the Kanban system and the concurrent ordering system. The effectiveness of the proposed system is clarified by the simulation experiments under various unstable demand conditions. As a result of simulation experiments, it is shown that the proposed controller can react to the unstable changes in demand, not only the mean but also the variance, with satisfying the required level for the mean waiting time of product demand. Furthermore, it can be shown that the proposed controller using neural network reduces the mean work-in-process inventories of the system much more than the previous controller using EWMA charts.